Envelope

Python module to extract the envelope of signals.

More information in this paper.

Interactive visualization here

Functions

This module implements 3 functions:

read_wav("path/to/mono/signal.wav"),

Returns a tuple (W, fps), where W is a numpy array and fps is an integer with the value of the frame rate of the file.

save_wav(signal, path = "test.wav", fps = 44100),

Saves a NumPy array as a .wav file.

get_frontiers(W, mode=0),

When mode=0:

Returns a tuple (pos_ids, neg_ids) of NumPy arrays, with the indices of the positive and negative frontiers of a NumPy array representing a discrete wave W.

When mode=1:

Returns the NumPy array X_envelope with the indices of the envelope of the wave W. Note that, to obtain the values of the envelope at those points, np.abs(W[X_envelope]) is recommended.

If this function is called from a compatible setup (currently, Windows 64bit), it computes the envelope and the frontiers faster via specialized native code. A fall-back version in pure Python is also provided, in which case a warning is printed indicating that a slower mode is being used. In both cases, the usage is the same, as are the results.

Usage

install the module: pip install signal-envelope

A minimal example would then be:

import signal_envelope as se

W, _ = se.read_wav("path/to/signal.wav")
X_pos_frontier, X_neg_frontier = se.get_frontiers(W, 0)
print(X_pos_frontier, W[X_pos_frontier])

X_envelope = se.get_frontiers(W, 1)
print(X_envelope, np.abs(W[X_envelope]))

A number of test wav files can be found at the Github repository for the project.

Source

The code for this repository is available at Github.

Used Libraries

Libsndfile Boost Intel Math Kernel Library